What the Confucian River Cannot See

Every strength is a corresponding blindspot. Every fishbowl reveals part of the world and hides the rest. The Confucian River's relational ontology — its greatest asset — is also the source of certain questions it has structural difficulty asking. To take the river seriously is to take its shadow seriously. This chapter names the shadow.

The first shadow concerns the individual whose relationships are constituted by oppressive structures.

The relational ontology says the person is constituted by relationships. The autonomous bounded individual is an abstraction; the real person is a node in a network. This is, in the cases where the relationships are healthy, a deepening of the conception of the human. It is also, in the cases where the relationships are not healthy, a structural difficulty.

Consider the woman in a Confucian family who is expected, by the cultivated norms of relational obligation, to subordinate her career to her husband's, her preferences to her in-laws', her voice to the consensus of the elders. The Confucian frame can name what is happening — she is fulfilling the obligations of a wife and daughter-in-law in a configuration of relationships that constitute her — but the frame has structural difficulty registering the configuration as oppressive in the way the liberal frame registers it. The liberal frame says: she has rights, prior to the configuration, that the configuration is violating. The Confucian frame says: she is becoming a person through the fulfillment of her relational obligations, and the cultivation of ren requires the acceptance of the burdens the relationships impose.

This is a real difference. It is not a difference that resolves cleanly. The liberal frame, in its excess, treats every relational obligation as a potential infringement on the prior individual autonomy that is the bedrock value. The Confucian frame, in its excess, treats every refusal of relational obligation as a deficiency of cultivation. Both excesses have produced real harm. The Western tradition of family law has, in its individualist enthusiasm, sometimes destroyed forms of family life that were actually flourishing. The Chinese tradition of family obligation has, in its relational enthusiasm, sometimes legitimated the suppression of individuals — particularly women, particularly younger sons, particularly anyone whose flourishing required the renegotiation of inherited relational structures.

The reform movements within the Chinese tradition have, of course, addressed this. The May Fourth generation took up the question of women's emancipation and the reform of family structures. The contemporary Chinese state has, in some domains, expanded individual rights beyond what the imperial tradition would have recognized. The Confucian frame is not static; it has been re-articulated, repeatedly, by people working within the tradition to address the structural injustices the tradition's relational emphasis has, at various points, legitimated.

But the underlying structural difficulty remains. The relational ontology, as a foundational philosophical commitment, has a harder time generating critical purchase against oppressive relational structures than the liberal individualist alternative. The liberal alternative has its own difficulties — it has trouble generating purchase against the dissolution of relational structures that are actually generative — but the difficulties run in opposite directions, and the Confucian River's difficulty is real.

For AI, the implication is specific. A Confucian AI ethics framework that evaluates AI systems by their effects on the relational fabric will have an easier time naming the harms of technologies that dissolve relational obligations and a harder time naming the harms of technologies that reinforce oppressive ones. An AI deployment that strengthens a hierarchical family structure in which one member is being subordinated to the others might pass the Confucian evaluation more easily than it would pass a liberal one. The strength of the Confucian frame in identifying technologies that erode community is also the weakness of the frame in identifying technologies that consolidate dominance within communities.

The second shadow concerns the minority whose flourishing requires the disruption of the cultivated consensus.

The Confucian frame, in its commitment to the substantive theory of the public good and the maintenance of wenming, has trouble with the figure who is right against the consensus. The dissident. The reformer. The minority whose voice the cultivated brightness has not yet learned to include. The Western liberal frame, with its commitment to individual rights and procedural protections for dissent, has more developed mechanisms for protecting the dissenter from the majority. The Confucian frame has historically dealt with the dissenter by treating dissent as a sign of insufficient cultivation, of failure to perceive the wisdom of the consensus, of incomplete moral development.

This is not the whole story. The Confucian tradition has, in its own register, valued the junzi (gentleman, exemplary person) who has the moral courage to advise the ruler against the consensus when the consensus is wrong. The remonstrating official is a recognized figure in Chinese political history. The Mencian insistence on the right of the people to overthrow a ruler who has lost the Mandate of Heaven is, in its own way, a doctrine of legitimate dissent.

But the figure who counts as a legitimate dissenter, in the Confucian frame, is the figure who is operating from inside the tradition and advancing what the tradition itself would, on reflection, recognize as the correct interpretation. The figure who is dissenting because they are outside the tradition — because the tradition has not yet recognized them as a full participant, because they belong to a minority whose perspective the cultivated brightness has not yet illuminated — has a harder time being heard.

For AI governance, the implication is again specific. The Chinese AI governance framework, with its substantive theory of the public good and its emphasis on outcome-based evaluation, has more difficulty protecting the AI deployment that benefits a minority at the expense of the consensus than the Western framework does. The Western framework, with its procedural protections for individual rights, can sometimes carve out spaces for minorities even against the consensus of the majority. The Chinese framework, oriented toward the maintenance of social harmony, will have a structural tendency to evaluate such deployments as disruptive.

This is not a hypothetical concern. The treatment of religious and ethnic minorities in contemporary China — Uyghur Muslims in Xinjiang, Tibetans, Falun Gong practitioners — has involved AI deployments that the substantive theory of the public good has legitimated and that the procedural rights framework would have, in principle, made harder to legitimate. The Confucian framework can register these deployments as disruptive of the relational fabric of the affected communities. The Confucian framework has more difficulty registering them as the violation of prior individual rights the rights tradition treats them as.

The third shadow concerns the asymmetry the Confucian frame has trouble surfacing between the cultivated consensus and the not-yet-cultivated.

The Confucian frame, in its emphasis on cultivation and the cultivation of wisdom, has a structural tendency to defer to those further along the path of cultivation. The teacher knows more than the student. The elder knows more than the younger. The cultivated knows more than the un-cultivated. The wisdom of the tradition is held more fully by those who have spent more time inside it.

This is, in many domains, simply true. The longer a serious person has spent inside a serious tradition, the more they tend to know. The Western academic tradition recognizes this as well — the senior scholar typically does know more than the graduate student. The difficulty arises when the deference becomes structural in a way that prevents the un-cultivated from being heard at all, and prevents the tradition from being corrected by perspectives the tradition has not yet recognized as legitimate.

For AI, the difficulty is this: the people most affected by the deployment of AI systems are not, typically, the people most cultivated in the tradition that is evaluating the deployment. The workers being displaced. The students being tutored. The patients being diagnosed. The citizens being surveilled. These are the populations on whom the technology is being deployed, and they are not, typically, the populations whose perspective the cultivated tradition is naturally inclined to weight.

The Western liberal tradition has procedural mechanisms for incorporating the perspective of the affected. Public consultation. Democratic deliberation. The right to be heard. These mechanisms are imperfect — often quite imperfect — but they are explicit, and they generate pressure that can sometimes counter the structural tendency of any tradition to defer to its own cultivated practitioners. The Confucian tradition's mechanisms for incorporating affected perspectives are less explicit. They depend on the cultivated officials being themselves committed to perceiving the situation accurately and acting in the public good. When the officials are committed and competent, the mechanism works. When they are not, the mechanism has fewer structural correctives than the liberal alternative.

This is not unique to China. Every tradition has the difficulty. The Western tradition's procedural mechanisms have, throughout history, failed to incorporate the perspectives of populations the tradition did not yet recognize — enslaved people, women, indigenous peoples, the colonized. The Western failures are at least as severe as the Confucian ones. But the procedural mechanisms, when they have worked, have provided a corrective the Confucian tradition has less developed equivalent of.

For AI governance, the implication is that the Chinese AI framework needs to develop more explicit mechanisms for incorporating the perspective of the populations being deployed on, especially when those populations are not well-represented in the cultivated official class that is doing the deploying. This is, in fact, an area of active reform within Chinese governance. But the structural difficulty remains, and the Confucian frame's relative weakness here is something the Confucian tradition has not yet fully addressed.

The fourth shadow is the deepest and the hardest to name. The Confucian tradition's substantive theory of the public good — its specification of what wenming requires, what social harmony looks like, what the cultivated civilization is supposed to be cultivating — is itself a substantive position. It is not a neutral starting point. It is a particular tradition's articulation of what flourishing looks like. The tradition's confidence in its own articulation, while in many cases earned, can become a structural barrier to recognizing that other articulations might be equally serious.

The liberal tradition has the inverse problem. It is so committed to neutrality among conceptions of the good that it has trouble even articulating what it itself is committed to. The Confucian tradition is so committed to its own substantive articulation that it has trouble taking the substantive articulations of other traditions as serious alternatives rather than as deviations from the cultivated norm.

For the global AI conversation, this means the Chinese voices in the conversation can have the difficulty their own framework has trouble naming: the assumption that the Confucian articulation of the public good is the correct substantive articulation, and that other civilizational traditions need to be brought into conversation with it as students of the cultivated wisdom the tradition has produced. This is not how Bing Song or Yi Zeng or Zheng Yongnian or Kai-Fu Lee or Robin Li would frame their work. None of them would claim the Confucian tradition has the universal answer. But the structural pressure of the tradition runs in that direction, and the global conversation has to be aware of the pressure even when the individual voices are explicitly resisting it.

The orange pill, in [YOU] on AI, makes the fishbowl visible. The Confucian River has a fishbowl too. The fishbowl is two and a half thousand years old. The fishbowl has produced some of the deepest insights humanity has on the relational structure of the human being and the cultivated conditions of civilizational flourishing. The fishbowl has also produced specific blindspots — the difficulty of seeing oppressive relational structures, the difficulty of protecting the minority against the consensus, the difficulty of incorporating the affected perspective when it is not represented in the cultivated class, the difficulty of recognizing other substantive articulations of the public good as equally serious.

These are real blindspots. Naming them is not a Western critique. It is a structural one. The strength and the shadow are the same. The relational ontology that produces the deepest insight into civilizational fabric also produces the structural difficulty in protecting the individuals whose flourishing requires the renegotiation of the fabric.

The Western tradition has the inverse pattern of strength and shadow. The individualist ontology that produces the deepest insight into individual dignity also produces the structural difficulty in maintaining the relational fabric in which the individual is constituted. Each tradition needs what the other supplies. Neither tradition can supply, by itself, what the situation requires.

That is the bridge. That is what Part V is going to build toward. Before we get there, we have to walk the other river, in equal length and equal seriousness, with the same orange-pill discipline of standing outside the assumptions long enough to see them as assumptions.

The Promethean River is waiting. Dry off. Step in. The water is faster and more recent and tastes different, and the stewards of the other river have their own structural commitments that produce their own structural insights and their own structural blindspots. Both are necessary. Neither is sufficient.

The water is moving. The bridge is being built. The view from the middle is the only view this century is going to permit.

Keep climbing.